Nothing like a ripe avocado

In the kitchen, determining the ripeness of an avocado is a simple matter of squeezing it to see how hard or squishy it is. Out in the orchard, however, determining whether avocados are ripe enough to be harvested is a more complex procedure, involving testing characteristics such as oil, moisture and dry matter contents. But it’s an important task, as avocados harvested before they’re ready can shrivel during storage and have a watery taste and rubbery texture, while those harvested too late have a short shelf-life and are more prone to disease.

At the moment, the oil, moisture and dry matter contents are determined using techniques that involve extraction and freezing, which destroy the avocado. As well as being time-consuming, these techniques can only test a subset of any harvest, giving just a general impression of its readiness and meaning that poor quality avocados will still get through. So Lembe Magwaza and his colleagues at the University of KwaZulu-Natal in South Africa decided to try replacing these destructive techniques with NIR spectroscopy.

During the 2013 and 2014 growing seasons, they collected a total of 155 avocados from a commercial farm, and then analysed a proportion of the intact fruit using NIR spectroscopy, focusing the NIR beam on a region around the middle of the fruit. This revealed that the most data-rich wavelengths were between 800nm and 2400nm, corresponding to stretching of the bonds between oxygen, hydrogen, carbon and water molecules. They then determined the oil, moisture and dry matter content of these avocados using conventional techniques.

Next, they used partial least squares regression to create various predictive models from the NIR data and the measured oil, moisture and dry matter contents. As reported in Scientia Horticulturae, when they tested these models on some of the other collected avocados, they found that the models for predicting moisture and dry matter content from NIR data were fairly accurate. However, the model for predicting oil content was much less accurate. Furthermore, models developed just from the avocados harvested in 2013 were not very accurate at predicting the contents for avocados harvested in 2014, demonstrating a strong seasonal effect, but models built from data for both 2013 and 2014 were much more accurate.

The inability of NIR spectroscopy to determine the oil content in avocados is not too much of a problem, as this can also be determined from the dry matter content. But Magwaza and his team are confident that the models can be made more accurate by including data from additional years.

Blog tags: 


dwhopkins's picture

Hi jonevans,
Has this work been published? I wonder whether the oil of the avocado tissue is actually being interrogated by the measurement beam?  If the light is not penetrating far enough below the skin, you won't be able to get a good correlation to the reference values for oil content.  It may be that the bulk moisture content is correlated with the moisture in the region of the skin, and you can get reasonable measurements of the moisture.  Of course the dry matter content is related to the moisture content, so you would also be able to obtain measurements of DM.  Would measurement of NIR interactance, where you detect the light reaching a detector placed some distance separated from the incident light beam, be able to see the deeper tissue and give you a better measurement of oil content?
Best wishes,
David Hopkins

venkynir's picture

Can you elborate the mode of measurment and source power . Why you could not try wavelet so that noise will be reduced
venkynir , India

ianm's picture

Venky and Dave:

Jon writes a couple of these blog posts for us each month; he is not the author of the research. The paper is linked from the blog post (click on Scientia Horticulturae in the penultimate para above). Probably best to take it up with the authors of the paper.

Do feel free to report what you find here at the end!

Best wishes